Can AI predict household spending? A scorecard across 6 surveys and 32 tasks
Most models beat a no-change baseline, but household predictions still carry systematic over- and under-prediction — and gradient-boosted trees still lead on most tasks.
Key numbers
- Benchmark scale: six US household surveys unified into 32 prediction tasks spanning numeric, categorical and probabilistic outcomes across consumption, income, labour, expectations and housing.
- Task design: predict household behaviour from past behaviour, demographics and macroeconomic conditions, including how households adjust to policy changes.
- Evaluation: 13 proprietary and open-weight LLMs against a no-change baseline and a gradient-boosted tree model; most LLMs beat the baseline, with the best model lowering error for numeric outcomes by 12.2%.
- Gradient-boosted trees rank first on most tasks, leading proprietary LLMs approach their performance while open-weight models lag, and LLMs show systematic over- and under-prediction across tasks.
- The paper also shows a reproducible fix: fine-tuning plus aggregating 16 predictions per observation lets a 4B-parameter open-weight model match proprietary models, generalising to policy-response tasks excluded from fine-tuning.
Method and sample
arXiv paper “HouseholdBench: Evaluating Large Language Models as Predictors of Household Economic Behavior” (2026-10-06): six US household surveys and 32 prediction tasks used to compare 13 models against a no-change baseline and gradient-boosted trees, including policy-response tasks and low-cost improvement methods.
What it means for brands
Two lessons for consumer and household brands: AI is not a better market-research model than established statistical methods, so “the AI said so” is not evidence; and the public facts it reads shape how it describes your category and your customers.
How SourceSight handles it
For consumer projects SourceSight builds prompts from purchase scenarios (which to choose, compare, pricing, is it right for me), reports mentions and position per engine, and uses the off-site list to cover review and community sources.
Limits
The benchmark is built on US household surveys and does not transfer directly to other markets; it measures behavioural prediction accuracy, not brand visibility in AI.
References
Run it on your own project
What the same prompts return across 10 engines is only knowable by running them.
Related research
Cross-category recommendations: +36% add-to-cart on item pages, +15% on cart pages
Recommending general merchandise to grocery shoppers — a milk frother with the milk — has long been hard; LLM retrieval plus real-time cart context moved production A/B metrics.
Industries · 2026-10-07Home services: 22% of homeowners already use AI to find a provider, and 80% of businesses don't know how to be found
The local services funnel is moving upstream: before a homeowner opens a website or calls, AI has already synthesised reviews, reputation and service descriptions into one answer.